The aim of this study was to examine the existing model approach and to make recommendations for improved modelling of under-ice dissolved oxygen (DO) concentrations in the Athabasca River. Previous modelling efforts have involved using the DO simulation model DOSTOC, but discrepancies between observed and predicted DO concentrations for 1992 and 1993 highlighted the needs for a review and assessment of the modelling approaches used. This report reviews the structure of the DOSTOC model and input data used for modelling, then describes a stochastic run of the model using the best available values for input data and process rates, including revised rates and ratios for converting five-day biological oxygen demand (BOD) values to ultimate BOD values, measured values for sediment oxygen demand and sedimentation, and a temperature correction for pulp mill BOD decay rates. Simulation results indicate whether the model could predict DO for 1990-94 winters. Regression equations were also developed relating DO concentration to river distance.